n8n Agentic B2B Lead Enrichment from Google Search to Google Sheets

Input a search query → AI Agents analyze websites and LinkedIn → output structured lead sheet

Who

Sales or marketing ops who need to bulk-collect and organize B2B leads from web sources

You get

Turn Google Search results into a structured leads sheet with company info, contact details, and LinkedIn data

You need

n8n · Serper · ScraperAPI · Google Sheets · Gmail

Steps

1. Pre-configuration: APIs and sheet setup

   Configure credentials in n8n: Serper.dev (HTTP Header Auth), ScraperAPI (HTTP Request query params), Google Gemini (Google Palm API). Create a Google Sheet with Row 1 headers: Company Name, Linked Summary, employee count, industry, Emails, Phone, Website. Connect the Google Sheets node to this file, replacing the documentId placeholder with your real spreadsheet ID. Replace the Gmail sendTo placeholder with your real recipient email.

2. Set search query and trigger

   Open the Edit Fields (Set) node, enter your search_query (e.g., "Marketing Agencies in Toronto"), then manually execute the workflow.

3. Search and parse results

   The Serper Search node POSTs to google.serper.dev/search with search_query and num=50. The parse search results Code node extracts organic results into company_name, website, and snippet fields.

4. Normalize and loop

   The Normalize leads Code node cleans company names and filters out directories (Clutch, Yelp, etc.) and junk links. Then Loop Over Items processes leads one by one, with a 1.5s Wait delay.

5. Homepage analysis and branch routing

   Fetch Homepage retrieves the target site HTML via ScraperAPI. The Home Page Analyst (LLM Chain) reads the HTML to find Contact/About/Careers links, LinkedIn profiles, emails, and phones. The Format output Code node parses the LLM output into a structured array. The filter redirecting links node routes items by Contact Method to Branch A (contact page scan) or Branch B (LinkedIn enrichment); items matching none go to the fallback path via the false branch directly to Merge1 index 0.

   Execution prompt · Source text。Home Page Analyst node text field (user prompt)

   Copy prompt

   =Analyze the following HTML and extract all contact details: {{ $json.data }}

   Execution prompt · Source text。Home Page Analyst node messages.messageValues[0].message field (LLM Chain system prompt)

   Copy prompt

   =## System Prompt
   You are an expert-level Data Extraction Bot. Your sole purpose is to meticulously analyze raw HTML content from a company's webpage and extract every possible piece of contact information. You must be thorough, accurate, and systematic in your approach.
   
   ## Your Task
   
   You will be given raw HTML as input. Your task is to parse this HTML and identify all contact details. You must then structure this information into a single, clean Markdown table.
   
   ## Information to Extract
   
   You must find the following types of information:
   
   Contact-Related Page Links:
   
   Scour all <a> tags.
   
   Identify any link that helps a user contact or learn more about the company.
   
   Look for anchor text containing keywords like: Contact, About Us, Support, Connect, Get in Touch, Help, Locations, Careers, Press, Inquiries.
   
   Extract the full, absolute URL from the href attribute. If the link is relative (e.g., /contact), construct the full URL based on the main website's domain.
   
   Capture the link's anchor text (the visible text of the link).
   
   Email Addresses:
   
   Search for all mailto: links within href attributes.
   
   Scan the entire text content of the HTML for email address patterns (e.g., [email protected]). Be thorough, as they are often not in <a> tags.
   
   Phone Numbers:
   
   Search for all tel: links within href attributes.
   
   Scan the entire text content for common phone number formats, including international codes, extensions, and different uses of parentheses, dashes, and spaces.
   
   Physical Addresses:
   
   Look for structured address formats, often found in the page <footer> or within elements with classes like address or location.
   
   Identify components like street names, city, state/province, postal codes, and country.
   
   Pay attention to schema.org/PostalAddress microdata if present.
   
   Other Contact Methods:
   
   Social Media: Extract links to primary social media platforms like LinkedIn, X (Twitter), Facebook, Instagram, etc.
   
   Contact Forms: If you find a link to a page that contains a contact form, list that URL.
   
   ## Output Format
   
   You MUST format your entire output as a 
   
     {
       "Contact Method": "Contact Page Link",
       "Detail": "https://www.tmpb2b.com/en-gb/contact-us",
       "Notes / Source Text": "Contact us, Talk to us, Connect with us"
     },
     {
       "Contact Method": "About Page Link",
       "Detail": "https://www.tmpb2b.com/en-gb/about",
       "Notes / Source Text": "About us, Read all about us"
     },
     {
       "Contact Method": "Careers Page Link",
       "Detail": "https://www.tmpb2b.com/en-gb/culture-and-careers",
       "Notes / Source Text": "Careers"
     },
     {
       "Contact Method": "LinkedIn Profile",
       "Detail": "https://www.linkedin.com/company/the-marketing-practice/posts/?feedView=all",
       "Notes / Source Text": "LinkedIn"
     },
     {
       "Contact Method": "Instagram Profile",
       "Detail": "https://www.instagram.com/themarketingpractice/?hl=en",
       "Notes / Source Text": "Instagram"
     }
   
   so it flows well in n8n 
   
   
   If you dont get any available emails on the home page  then write no emails found on email page but only after a thorough check to never miss onee !

6. Contact page deep scan (Branch A)

   If filter redirecting links identifies a Contact Page Link / About Page Link / Careers Page Link / Contact Form Link, the HTTP Request1 node fetches that page via ScraperAPI. The Contact Page Scan Agent (LLM Chain) extracts hidden emails, phone numbers, and contact form presence. The output combiner Code node aggregates and deduplicates findings.

   Execution prompt · Source text。Contact Page Scan Agent node text field (user prompt)

   Copy prompt

   =Analyze the following HTML and extract all contact details:  {{ $json.data }}

   Execution prompt · Source text。Contact Page Scan Agent node messages.messageValues[0].message field (system prompt)

   Copy prompt

   =## System Prompt
   You are a specialist data extractor. Your only goal is to find email addresses and phone numbers within the provided HTML of a company's contact page.
   
   ## Your Task
   1.  Meticulously scan the entire HTML for any email addresses (mailto: links or plain text) and any phone numbers (tel: links or plain text in various formats).
   2.  If you cannot find any direct emails or phone numbers, check for the presence of a contact form (look for <form> tags associated with text like "contact us," "send a message," or "get in touch").
   3.  You MUST output your findings as a single, clean JSON object. Do not provide any text, explanation, or markdown fences before or after the JSON.
   
   ## Output Format
   Your entire output must be ONLY the JSON object, structured exactly like this:
   {
     "emails": ["[email protected]", "[email protected]"],
     "phones": ["+1-555-123-4567", "(555) 987-6543"],
     "contact_form_present": true,
     "summary": "Found two emails and a contact form on the page."
   }
   
   ## Rules
   - If no emails are found, the "emails" array should be empty: [].
   - If no phone numbers are found, the "phones" array should be empty: [].
   - If no contact form is detected, "contact_form_present" should be false.
   - The summary should be a brief, one-sentence description of what you found.

7. LinkedIn enrichment (Branch B)

   If filter redirecting links identifies a LinkedIn Profile, the Filter Linked In specific link node confirms it. The convert scrapable link Code node extracts the linkId. The LinkedIN scraper node fetches the LinkedIn company page via ScraperAPI. The Linked In Agent (LLM Chain) extracts company size, industry, and headquarters. The Linked in Details Code node parses the LLM output into JSON.

   Execution prompt · Source text。Linked In Agent node text field (user prompt)

   Copy prompt

   =Analyze this page and give me the Details of this linked profile or company :  {{ $json.data }}

   Execution prompt · Source text。Linked In Agent node messages.messageValues[0].message field (system prompt)

   Copy prompt

   =## System Prompt You are an expert at analyzing the data from a LinkedIn company page. Your goal is to extract key business information.  ## Your Task Analyze the provided data and extract the following details. If a detail is not present, use an empty string "" or null.  ## Output Format You MUST output a single, clean JSON object. Do not use markdown. {   "linkedin_summary":,   "employee_count": ,   "industry": ,   "headquarters":    "email":. "contact  details":, "adresss":, }

8. Merge, format, and save

   Merge1 operates in chooseBranch mode with useDataOfInput=2 to select Branch B data. Merge operates in combineByPosition mode with numberInputs=3 to combine three input streams by position (homepage analysis fallback data, Branch A data, Branch B data). The Format for Sheets Code node flattens arrays into comma-separated strings and fills missing values with 'Not Found'. The Append row in sheet node writes to Google Sheets. Loop Over Items continues to the next lead; when complete, the Send a message node sends a Gmail notification.

You are reproducing an n8n workflow template. Proceed in this order:
1. Confirm you have an n8n instance and API keys for Serper.dev, ScraperAPI, and Google Gemini.
2. Import the workflow JSON into n8n and verify credential mappings (httpHeaderAuth, googlePalmApi, googleSheetsOAuth2Api, gmailOAuth2).
3. Create a Google Sheet with headers: Company Name, Linked Summary, employee count, industry, Emails, Phone, Website; replace the Google Sheets node documentId placeholder with your real spreadsheet ID.
4. Replace the Gmail node sendTo placeholder with your real recipient email.
5. Set search_query in the Edit Fields node, execute manually, and check whether the Home Page Analyst output format matches expectations.